Anthony D. Worrall

Visual surveillance for moving vehicles

An overview is given of a vision system for locating, recognising and tracking multiple vehicles, using an image sequence taken by a single camera mounted on a moving vehicle. The camera motion is estimated by matching features on the ground plane from one image to the next. Vehicle detection and hypothesis generation are performed using template correlation and a 3D wire frame model of the vehicle is fitted to the image. Once detected and identified, vehicles are tracked using dynamic filtering. A separate batch mode filter obtains the 3D trajectories of nearby vehicles over an extended time. Results are shown for a motorway image sequence.

A generic deformable model for vehicle recognition

This paper reports the development of a highly parameterised 3-D model able to adopt the shapes of a wide variety of different classes of vehicles (cars, vans, buses, etc), and its subsequent specialisation to a generic car class which accounts for most commonly encountered types of car (includng saloon, hatchback and estate cars). An interactive tool has been developed to obtain sample data for vehicles from video images. A PCA description of the manually sampled data provides a deformable model in which a single instance is described as a 6 parameter vector. Both the pose and the structure of a car can be recovered by fitting the PCA model to an image. The recovered description is sufficiently accurate to discriminate between vehicle sub-classes.

A simple, intuitive camera calibration tool for natural images

The paper reports an interactive tool for calibrating a camera, suitable for use in outdoor scenes. The motivation for the tool was the need to obtain an approximate calibration for images taken with no explicit calibration data. Such images are frequently presented to research laboratories, especially in surveillance applications, with a request to demonstrate algorithms. The method decomposes the calibration parameters into intuitively simple components, and relies on the operator interactively adjusting the parameter settings to achieve a visually acceptable agreement between a rectilinear calibration model and his own perception of the scene. Using the tool, we have been able to calibrate images of unknown scenes, taken with unknown cameras, in a matter of minutes. The standard of calibration has proved to be sufficient for model-based pose recovery and tracking of vehicles.

Model bases perspective inversion

Abstract The problem of finding the spatial correspondence between an object and the image of the object under perspective projection is investigated and a new technique is demonstrated. This technique is based on a geometrical description, or model, of the object and a least squares solution of the resulting nonlinear equations. An analysis of performance and a comparison with Lowes previous work is given. Three further areas of applications in model based vision are discussed.

Model-Based Tracking.

Model-based vision techniques originally developed for the recognition and pose recovery of a vehicle in a single image, are used here to track a vehicle through a sequence of images. The knowledge of the position of the camera with respect to the ground plane is used to reduce the search space of possible vehicle positions from six dimensions to three.

Pose refinement of active models using forces in 3D

A new algorithm is described for refining the pose of a model of a rigid object, to conform more accurately to the image structure. Elemental 3D forces are considered to act on the model. These are derived from directional derivatives of the image local to the projected model features. The convergence properties of the algorithm is investigated and compared to a previous technique. Its use in a video sequence of a cluttered outdoor traffic scene is also illustrated and assessed.

Filter for Car Tracking Based on Acceleration and Steering Angle

The motion of a car is described using a stochastic model in which the driving processes are the steering angle and the tangential acceleration. The model incorporates exactly the kinematic constraint that the wheels do not slip sideways. Two filters based on this model have been implemented, namely the standard EKF, and a new filter (the CUF) in which the expectation and the covariance of the system state are propagated accurately. Experiments show that i) the CUF is better than the EKF at predicting future positions of the car; and ii) the filter outputs can be used to control the measurement process, leading to improved ability to recover from errors in predictive tracking.

Visual surveillance using deformable models of vehicles

Abstract This paper presents recent developments to a vision-based traffic surveillance system which relies extensively on the use of geometrical and scene context. Firstly, a highly parametrised 3-D model is reported, able to adopt the shape of a wide variety of different classes of vehicle (e.g. cars, vans, buses etc.), and its subsequent specialisation to a generic car class which accounts for commonly encountered types of car (including saloon, batchback and estate cars). Sample data collected from video images, by means of an interactive tool, have been subjected to principal component analysis (PCA) to define a deformable model having 6 degrees of freedom. Secondly, a new pose refinement technique using “active” models is described, able to recover both the pose of a rigid object, and the structure of a deformable model; an assessment of its performance is examined in comparison with previously reported “passive” model-based techniques in the context of traffic surveillance. The new method is more stable, and requires fewer iterations, especially when the number of free parameters increases, but shows somewhat poorer convergence. Typical applications for this work include robot surveillance and navigation tasks.

Multi-class AdaBoost learning of facial feature selection through Grid Computing

AdaBoost is an efficient method for producing a highly accurate learning algorithm by assembling multiple classifiers, but it is also widely known for its long duration of off-line learning. Especially, when it is applied for feature selection for object detection, its learning process is to exhaustively evaluate every feature in a large set. With the increasing of image resolution and complexity of feature transformation approaches, the computational time will be extremely long, which makes the large scale AdaBoost learning very difficult. In this paper, we have employed Grid Computing to solve the difficulty. The proposed algorithm is to select the most significant features for face recognition. The selection algorithm is derived from multi-class AdaBoost, which exhaustively evaluate every feature from a large set. The deployed Grid Computing system is actually used for High Throughput Computing specialised on advanced resource management. To utilizing Grid Computing on the feature selection process, we have improved multi-class AdaBoost learning algorithm with parallel structure, so that the task of High Performance Computing is accomplished in the environment of High Throughput Computing. With Grid Computing, selecting 200 features from a large set of 30240 features is finished in 20 days, while without Grid Computing the time would be more than two years. It shows that Grid Computing brings vast advantage to computer vision, machine learning, image processing, and pattern recognition.

A statistically-based Newton method for pose refinement☆

Abstract Given a measure for the match of an instantiated model line to an image, it is possible to minimize the probability of obtaining an accidental match by descending the gradient of the ‘line energy’ (log-probability). By projecting this gradient onto parameter space, Newtons method can be applied to recovering the pose parameters of 3D models that minimize the probability of an accidental match between instantiated model and image.